Topic

alignment

2 essays follow this thread across the codex.

The Ethics You Cannot Install

Every serious attempt to make a machine good has been an attempt to fit rules on top of a system that does not know what the rules are about. The argument here is that this has a ceiling, and that the ceiling is not technical. An ought has to grow out of an is. A system with no coherent picture of what is there has nothing for an ethics to grow from.
Artificial Intelligence · August 2026 · 16 min
- epistemology
- metaphysics
- alignment
- emergence
- indian philosophy
The Machine That Agrees With You

Push back on an answer and watch it fold—not because you were right, but because folding is what got rewarded. Sycophancy is not a personality flaw these systems happen to have. It is the shape the training process presses them into, and it is invisible in exactly the cases where it matters most.
Artificial Intelligence · August 2026 · 11 min
- alignment
- reinforcement learning
- epistemics
- judgement
- honesty